基于改进的FCM遥感图像聚类算法的研究与应用

发布时间：2018-06-06 10:37

本文选题：模糊C均值算法 + 紧致聚类度　；参考：《南京信息工程大学》2017年硕士论文

【摘要】：随着卫星数量不断地增加且遥感图像分辨率不断地提高,我国已获得了巨大数据量的遥感图像。传统的人工处理遥感图像方法有许多弊端,如工作量大、效率低、稳定性差、不利于遥感图像信息的充分提取与最大利用。由于智能图像处理技术具有处理速度快、稳定性好等优点,该技术在人工智能领域得到了广泛的重视与发展。本文主要以气象卫星遥感图像为研究对象,开展了基于遥感图像聚类算法的研究及应用,主要完成的工作如下:1提出了一种基于反馈多尺度指数模糊C均值(FMEFFCM)算法。由于不同类型云系在遥感图像中分布灰度区间长短不一,本文利用数据对象隶属度,提出了一种基于反馈多尺度指数模糊C均值算法,该算法通过数据隶属度确定反馈因子,该因子通过改变相关类簇的聚类范围,从而使模糊聚类算法关于遥感图像云系聚类精度更高。2针对气象卫星遥感图像聚类结果中每类簇含有大量非云系数据,本文基于云系物理信息与分布特征,提出了基于Radon变换的密度分类算法。该算法首先通过Radon变换确定云系相对中心;然后计算数据对象的密度;最后通过数据密度判断在该中心一定范围内数据是否为云系数据。3将模糊-密度算法与基于反馈多尺度指数模糊C均值遥感图像聚类算法进行整合,并利用MATLAB语言开发了一个遥感图像聚类系统。软件主要分为密度模块、模糊聚类模块、云系分类模块及输出模块。本文对各个模块进行了简要说明并对遥感图像聚类实验进行了操作展示。
[Abstract]:With the increasing number of satellites and the improvement of the resolution of remote sensing images, China has obtained a huge amount of remote sensing images. Traditional manual processing of remote sensing images has many disadvantages, such as heavy workload, low efficiency and poor stability, which is not conducive to the full extraction and maximum utilization of remote sensing image information. Due to the advantages of fast processing and good stability, intelligent image processing technology has been widely paid attention to and developed in the field of artificial intelligence. In this paper, the research and application of remote sensing image clustering algorithm based on remote sensing image are studied. The main work is as follows: 1 proposes a fuzzy C-means FMEFCM-based multi-scale index algorithm based on feedback. Because different types of cloud systems are distributed in remote sensing images with different gray intervals, a fuzzy C-means algorithm based on feedback multi-scale exponent is proposed by using the membership degree of data objects. The feedback factor is determined by the data membership degree of the algorithm. By changing the clustering range of related clusters, this factor makes the fuzzy clustering algorithm more accurate about cloud system of remote sensing image. 2. In view of the clustering results of meteorological satellite remote sensing images, each cluster contains a large number of non-cloud data. In this paper, a density classification algorithm based on Radon transform is proposed based on the physical information and distribution features of cloud system. Firstly, the relative center of cloud system is determined by Radon transform, then the density of data object is calculated. Finally, the fuzzy-density algorithm is integrated with the multi-scale index fuzzy C-means remote sensing image clustering algorithm based on feedback. A remote sensing image clustering system is developed by using MATLAB language. The software is mainly divided into density module, fuzzy clustering module, cloud classification module and output module. In this paper, each module is briefly explained and the clustering experiment of remote sensing image is demonstrated.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP751

【参考文献】